typically I think "balooning" occurs when VMs have memory that they are not actively using AND there are other VMs that are requesting/needing more memory. So the host asks vmware tools to "baloon" so that the vm thinks it still has active physical memory, but really a bunch of it is from disk swap space. That frees up actual physical memory that the host can then assign to the other vm.
I think there is balooning in this situation because the limit is so much lower than the configured - and vmtools is "balooning" so that the vm thinks it has what it was configured with. These are contradictory settings, so ... I don't think it is optimal.